Spring onset variability and trends in the CMIP5 archive (Invited)

Here we characterize 'spring onset' using a well-vetted and representative model of plant leaf-out (the Leaf Index developed by M. D. Schwartz). The model relies only on atmospheric variables (daily Tmin and Tmax) as well as day length as a function of latitude. This phenological model has been used as an index to track regional to continental-scale variations in the timing of spring over the last century, relate them to land surface phenology, and explore their climatologies. Here, we calculate this index from climate reanalysis data and from daily Climate Model Intercomparison 5 (CMIP5) output to assess predictability of spring onset in a changing climate. We found that trends toward earlier spring in observations are nearly a factor of two higher than those simulated using the CMIP5 archive from 1979-2005 for most continental Northern Hemisphere regions (~1 days/decade as opposed to 0.5). Moreover regional patterns do not agree between models and observations, or even between climate models. At face value, this indicates a failure of the climate models to represent the 'true' patterns and trends in the timing of spring. An alternative explanation is that much of the regional variability in observed spring onset on decadal timescales is paced by internal climate fluctuations not captured in the model runs. This interpretation is supported by results obtained from a large ensemble of simulations from a single model (CCSM4), which show as much regional variability as the multi-model ensemble of CMIP5 simulations over the period 1979-2005. Further work should focus on comparing these and other climate-based indices of spring onset with climate model phenology, as well as diagnosing the climatic processes responsible for interannual and decadal variations in spring onset.